Package BCRA. April 25, 2018

Type Package Title Breast Cancer Risk Assessment Version 2.0 Date 2018-04-25 Author Fanni Zhang Maintainer Fanni Zhang <zhangf@imsweb.com> Package BCRA April 25, 2018 Functions provide risk projections of invasive breast cancer based on Gail model according to National Cancer Institute's Breast Cancer Risk Assessment Tool algorithm for specified race/ethnic groups and age intervals. License GPL (>= 2) NeedsCompilation no Repository CRAN Date/Publication 2018-04-25 15:35:14 UTC R topics documented: BCRA-package....................................... 2 absolute.risk......................................... 3 BrCa_1_AR......................................... 4 BrCa_beta.......................................... 5 BrCa_lambda1....................................... 5 BrCa_lambda2....................................... 6 check.summary....................................... 7 error.table.......................................... 8 error.table.all........................................ 8 exampledata......................................... 9 list.constants......................................... 10 recode.check........................................ 11 relative.risk......................................... 13 risk.summary........................................ 14 Index 16 1

2 BCRA-package BCRA-package A Package for Breast Cancer Risk Assessment Details This package is to project absolute risk of invasive breast cancer according to NCI s Breast Cancer Risk Assessment Tool (BCRAT) algorithm for specified race/ethnic groups and age intervals. The updated version 2.0 includes the new Hispanic model. This package can be used to estimate the risk of developing breast cancer over a predetermined time interval with risk factors. As the same as Breast Cancer Risk Assessment SAS Macro, the users can specify the time interval as appropriate, not only limited to the 5 years risk prediction available with BCRAT. The main function in this package is absolute.risk, which is defined based on a statistical model known as the "Gail model". Parameters and constants needed in this function include initial and projection age, recoded covariates using function recode.check, relative risks of BrCa at age "<50" and ">=50" obtained from function relative.risk as well as other known constants listed from function list.constants like BrCa composite incidences, competing hazards, 1-attributable risk using in NCI BrCa Risk Assessment Tool (NCI BCRAT). With risk factors and projection interval ages for a group of women, the function absolute.risk will return the corresponding absolute risk projections. If the function returns any missing values, the function error.table or error.table.all is used to find where the errors occured. The function check.summary can give a quick check for errors of input file and missing values of risks. For further analysis, a data frame is created from the function risk.summary, which includes age, duration of the projection time interval, covariates and the projected risk. The version 2.0 includes absolute risk projections for Hispanic women (US born and Foreign born) based on race specific RR risk models developed on the San Francisco Bay Area Breast Cancer Study (SFBCS). Race specific attributable risks, breast cancer composite incidences and competing hazards are added to the updated package. Author(s) Fanni Zhang <zhangf@imsweb.com> References Banegas MP, John EM, Slattery ML, Gomez SL, Yu M, LaCroix AZ, Pee D, Chlebowski RT, Hines LM, Thompson CA, Gail MH. Projecting Individualized Absolute Invasive Breast Cancer Risk in US Hispanic Women. JNCI 2016; 109. Matsuno RK, Costantino JP, Ziegler RG, Anderson GL, Li H, Pee D, Gail MH. Projecting individualized absolute invasive breast cancer risk in asian and pacific islander american women. JNCI 103(12):951-61, 2011. Gail MH, Costantino JP, Pee D, Bondy M, Newman L, Selvan M, Anderson GL, Malone KE, Marchbanks PA, McCaskill-Stevens W, Norman SA, Simon MS, Spirtas R, Ursin G, Berstein L.

absolute.risk 3 Projecting individualized absolute invasive breast cancer risk in African American women. JNCI 99(23):1782-92, 2007. Costantino J, Gail MH, Pee D, Anderson S, Redmond CK, Benichou J, Wieand HS. Validation studies for models to project the risk of invasive and total breast cancer. JNCI 91(18):1541-48, 1999. Gail MH, Brinton LA, Byar DP, Corle DK, Green SB, Shairer C, Mulvihill JJ. Projecting individualized probabilities of developing breast cancer for white females who are being examined annually. JNCI 81(24): 1879-86, 1989. absolute.risk Estimate absolute risks A function to estimate absolute risks of developing breast cancer absolute.risk(data, Raw_Ind=1, Avg_White=0) Arguments data Raw_Ind Avg_White A data set containing all the required input data needed to perform risk projections, such as initial age, projection age, BrCa relative risk covariates and race. See exampledata for details. The raw file indicator with default value 1. Raw_Ind=1 means RR covariates are in raw/original format. Raw_Ind=0 means RR covariates have already been re-coded to 0, 1, 2 or 3. Calculation indicator. Avg_White=0, calculate absolute risks; Avg_White=1, calculate average absolute risks based on the rates for average non-hispanic white women and average other (native american) women. The default value is 0. Details For the projection of absolute risks, this function is defined based on Gail Model. Parameters and constants needed in this function include initial and projection age, recoded covariates from function recode.check, relative risks of BrCa at age "<50" and ">=50" from function relative.risk as well as other known constants like BrCa composite incidences, competing hazards, 1-attributable risk using in NCI BrCa Risk Assessment Tool (NCI BCRAT). Value A vector which returns absolute risk values when Avg_White=0 or average absolute risk values when Avg_White=1.

4 BrCa_1_AR See Also recode.check, relative.risk Examples data(exampledata) # calculate absolute risk absolute.risk(exampledata) # calculate average absolute risk Avg_White <- 1 absolute.risk(exampledata, Raw_Ind=1, Avg_White) BrCa_1_AR Breast cancer 1-Attributable Risk 1-Attributable Risk data("brca_1_ar") Format A data frame with 2 observations on the following 5 variables. Wh.Gail White AA.CARE African-American HU.Gail Hispanic-American (US born) NA.Gail Other (Native American and unknown race) HF.Gail Hispanic-American (Foreign born) Asian.AABCS Asian-American

BrCa_beta 5 BrCa_beta Breast cancer beta The logistic regression coefficients derived from the Gail model. data("brca_beta") Format A data frame with 6 observations on the following 5 variables. Wh.Gail White, Gail model AA.CARE African-American, Care model HU.Gail Hispanic-American (US born), Gail model NA.Gail Other (Native American and unknown race), Gail model HF.Gail Hispanic-American (Foreign born), Gail model Asian.AABCS Asian-American, AABCS model BrCa_lambda1 Breast cancer composite incidences Breast cancer composite incidences for different races and age groups from 20 to 90 by 5 years. data("brca_lambda1") Format A data frame with 14 age groups on the following 12 variables. Wh.1983_87 White SEER 1983:1987 AA.1994_98 African-American SEER 1994:1998 HU.1995_04 Hispanic-American (US born) 1995:2004 NA.1983_87 Native American and unknown race 1983:1987 HF.1995_04 Hispanic-American (Foreign born) 1995:2004 Ch.1998_02 Chinese-American SEER 18 1998:2002

6 BrCa_lambda2 Ja.1998_02 Japanese-American SEER 18 1998:2002 Fi.1998_02 Filipino-American SEER 18 1998:2002 Hw.1998_02 Hawaiian SEER 18 1998:2002 op.1998_02 Other Pacific Islander SEER 18 1998:2002 oa.1998_02 Other Asian SEER 1998:2002 Wh_Avg.1992_96 Average White SEER 1992:1996 BrCa_lambda2 Breast cancer competing mortality Breast cancer competing mortality for different races and age groups from 20 to 90 by 5 years. data("brca_lambda2") Format A data frame with 14 age groups on the following 12 variables. Wh.1983_87 White SEER 1983:1987 AA.1994_98 African-American SEER 1994:1998 HU.1995_04 Hispanic-American (US born) 1995:2004 NA.1983_87 Native American and unknown race 1983:1987 HF.1995_04 Hispanic-American (Foreign born) 1995:2004 Ch.1998_02 Chinese-American SEER 18 1998:2002 Ja.1998_02 Japanese-American SEER 18 1998:2002 Fi.1998_02 Filipino-American SEER 18 1998:2002 Hw.1998_02 Hawaiian SEER 18 1998:2002 op.1998_02 Other Pacific Islander SEER 18 1998:2002 oa.1998_02 Other Asian SEER 1998:2002 Wh_Avg.1992_96 Average White SEER 1992:1996

check.summary 7 check.summary Summarize the error indicators, relative risks and absolute risks A function to show descriptive statistics by applying function mean and sd to the quantities Error_Ind, AbsRisk, RR_Star1 and RR_Star2. check.summary(data, Raw_Ind=1, Avg_White=0) Arguments data Raw_Ind Avg_White A data set containing all the required input data needed to perform risk projections, such as initial age, projection age, BrCa relative risk covariates and race. See exampledata for details. The raw file indicator with default value 1. Raw_Ind=1 means RR covariates are in raw/original format. Raw_Ind=0 means RR covariates have already been re-coded to 0, 1, 2 or 3. Calculation indicator. Avg_White=0, calculate absolute risks; Avg_White=1, calculate average absolute risks based on the rates for average non-hispanic white women and average other (native american) women. The default value is 0. Details When the mean and standard deviation for the variable Error_Ind is 0, implies that no errors have not been found. Otherwise when the mean and std for Error_Ind is not 0, implies that errors have been found. When errors are found, the number of records with errors is the count asscociated with AbsRisk listed under NMiss (number of missing). Value A summary table for error indicators, relative risks and absolute risks See Also recode.check, relative.risk, absolute.risk

8 error.table.all error.table List the records and errors for IDs with missing absolute risks A function to list the records and errors for IDs with missing absolute risks. For each of the records with error, the record is listed followed by a line which gives some indication as to where the error occured. Relative risks and risk pattern numbers are also included. error.table(data, Raw_Ind=1) Arguments data Raw_Ind A data set containing all the required input data needed to perform risk projections, such as initial age, projection age, BrCa relative risk covariates and race. See exampledata for details. The raw file indicator with default value 1. Raw_Ind=1 means RR covariates are in raw/original format. Raw_Ind=0 means RR covariates have already been re-coded to 0, 1, 2 or 3. Value A data frame listing the raw records, errors, relative risks and pattern numbers for IDs with missing absolute risks. If there is nothing wrong with the input data, the function will return "NO ERROR!". See Also recode.check, error.table.all error.table.all List all records and errors A function to list all records with both raw values and recoded values (or indications for errors). For each of the records, the record is listed followed by a line which gives some indication as to where the error occured. error.table.all(data, Raw_Ind=1)

exampledata 9 Arguments data Raw_Ind A data set containing all the required input data needed to perform risk projections, such as initial age, projection age, BrCa relative risk covariates and race. See exampledata for details. The raw file indicator with default value 1. Raw_Ind=1 means RR covariates are in raw/original format. Raw_Ind=0 means RR covariates have already been re-coded to 0, 1, 2 or 3. Value A data frame listing all records and errors. If there is nothing wrong with the input data, the function will return "NO ERROR!". See Also recode.check, error.table exampledata Example data set A data set containing all the required input data needed to perform risk projections, such as initial age, projection age, BrCa relative risk covariates and race. data("exampledata") Format Details A data frame with 26 observations on the following 9 variables. ID Woman s ID, positive integer 1, 2, 3,... T1 Initial age, all real numbers T1 in [20, 90). T2 BrCa projection age, all real numbers T2 in (20,90] such that T1<T2. N_Biop The number of biopsies, 0, 1, 2,..., 99=unk (99 recoded to 0). HypPlas Did biopsy display atypical hyperplasia? 0=no, 1=yes, 99=unk or not applicable. AgeMen Age at menarchy, less than or equal to initial age, 99=unk. Age1st Age at first live birth, greater or equal to age at menarchy and less than or equal to initial age, 98=nulliparous, 99=unk. N_Rels The number of 1st degree relatives with BrCa, 0, 1, 2,... 99=unk. Race Race, positive integer 1, 2, 3,...,11. See details.

10 list.constants 1=Wh White 1983-87 SEER rates (rates used in NCI BCRAT) 2=AA African-American 3=HU Hispanic-American (US born) 1995-04 4=NA Other (Native American and unknown race) 5=HF Hispanic-American (Foreign born) 1995-04 6=Ch Chinese-American 7=Ja Japanese-American 8=Fi Filipino-American 9=Hw Hawaiian-American 10=oP Other Pacific Islander 11=oA Other Asian list.constants List all constants required for BrCa absolute risk projections A function to create a text file under user s working directory which contains all constants required for BrCa absolute risk projections. list.constants(brca_lambda1, BrCa_lambda2, BrCa_beta, BrCa_1_AR) Arguments BrCa_lambda1 BrCa_lambda2 BrCa_beta BrCa_1_AR Breast Cancer Composite Incidences Breast Cancer Competing Mortality The logistic regression coefficients (beta) derived from the Gail model 1-Attributable Risk Details See "BrCa_lambda1.rda", "BrCa_lambda2.rda", "BrCa_beta.rda", "BrCa_1_AR.rda" in package data folder. Value A text file "list_all_constants.txt" exported under user s working directory for reading convenience.

recode.check 11 recode.check Recode and check the relative risk covariate values A function to recode the relative risk covariates and check errors. recode.check(data, Raw_Ind=1) Arguments data Raw_Ind A data set containing all the required input data needed to perform risk projections, such as initial age, projection age, BrCa relative risk covariates and race. See exampledata for details. The raw file indicator with default value 1. Raw_Ind=1 means RR covariates are in raw/original format. Raw_Ind=0 means RR covariates have already been re-coded to 0, 1, 2 or 3. Details This function is to recode the following relative risk covariates. Recoded RR covariates are named as NB_Cat, AM_Cat, AF_Cat and NR_Cat for N_Biop, AgeMen, Age1st and N_Rels, respectively. N_Biop: AgeMen: Age1st: N_Rels: The number of biopsies. Age at menarchy. Age at first live birth. The number of first degree relatives with BrCa. See the following table for recoding details. Covariate Raw Value Recoded to N_Biop 0 or 99 (unk or not applicable) 0 1 1 2,3,4... and not 99 2 AgeMen 14,15,16... or 99 (unk) 0 12,13 1 11 and younger 2 Age1st 19 and younger or 99 (unk) 0 20,21,22,23,24 1 25,26,27,28,29 or 98 (nulliparous) 2 30,31,32... and not 98 and not 99 3

12 recode.check N_Rels 0 or 99 (unk) 0 1 1 2,3,4... and not 99 2 This function is also used to check consistency and errors of input data. Let set_t1_missing and set_t2_missing be the checking variables for T1 and T2. The constraint on T1 and T2 is 20<=T1<T2<=90. If it is violated, set_t1_missing and set_t2_missing and the absolute risk will be set to the missing value NA. Let RacCat be the checking variable for Race. If the Race value is not included in the 11 races defined, the absolute risk will be set to the missing value NA and RacCat will be set to "U" (undefined). The corresponding character of Race CharRace will be set to "??". Let set_hyperp_missing and set_r_hyp_missing be the checking variables for HypPlas and R_Hyp. Consistency patterns for the number of Biopsies and Hyperplasia are: Requirment (A) N_Biops=0 or 99, then HypPlas MUST = 99 (not applicable). Requirment (B) N_Biops>0 and <99, then HypPlas = 0, 1 or 99. If ANY of the above 2 REQUIREMENTS is violated, NB_Cat, set_hyperp_missing and set_r_hyp_missing will be set to the corresponding character "A" or "B" and the absolute risk will be set to the missing value NA. The consequences to the relative risk (RR) for the above two requirements are: (A) N_Biops=0 or 99, HypPlas=99 (not applicable) inflates RR by 1.00. (B) N_Biops>0 and <99, HypPlas=0 (no) inflates RR by 0.93; N_Biops>0 and <99, HypPlas=1 (yes) inflates RR by 1.82; N_Biops>0 and <99, HypPlas=99 (unk) inflates RR by 1.00. For remaining relative risk covariates, AgeMen, Age1st and N_Rels: AgeMen Age at menarchy must be postive integer less than or equal to initial age T1. NOTE: (1) For African-American women AgeMen<=11 are grouped with AgeMen=12 or 13. (2) For US Born Hispanic women AgeMen is not included in the RR model and all values for this variable are recoded to 0. Age1st Age at 1st live birth must be postive integer greater than equal to AgeMen and less than or equal to initial age T1. NOTE: (1) For African-American women, Age1st is not included in the RR model and all values for this variable are recoded to 0. (2) For US Born and Foreign Born Hispanic women, the recoding for this variable follows: Age1st 19 and younger or 99 (unk) 0 20-29 1 30+ or 98 (nulliparous) and not 99 2 N_Rels The number of 1st degree relatives with BrCa must be 0,1,2...

relative.risk 13 NOTE: For Asian-Americans Race=6-11 and Hispanic-Americans (US and foreign born), the number of 1st degree relative coded value of 2 gets grouped with 1. Value A data frame containing the error indictors, recoded covariates as well as other checking variables defined for checking the consistency of the input data. See Also error.table.all, error.table Examples data(exampledata) recode.check(exampledata) relative.risk Estimate relative risks A function to estimate relative risks for risk factor combinations relative.risk(data, Raw_Ind=1) Arguments data Raw_Ind A data set containing all the required input data needed to perform risk projections, such as initial age, projection age, BrCa relative risk covariates and race. See exampledata for details. The raw file indicator with default value 1. Raw_Ind=1 means RR covariates are in raw/original format. Raw_Ind=0 means RR covariates have already been re-coded to 0, 1, 2 or 3. Details The age is dichotomized as "age less than 50 years" and "age 50 years or more". The relative risks can be obtained from Gail Model, an unconditional logistic regression that included main effects NB_Cat, AM_Cat, AF_Cat, NR_Cat as well as interactions between AF_Cat and NR_Cat and between the age category and NR_Cat.

14 risk.summary Value See Also RR_Star1 Relative risk for woman of interest at ages < 50. RR_Star2 Relative risk for woman of interest at ages >= 50. PatternNumber recode.check The sequence number of risk patterns. There are 3 levels for NB_Cat, 3 for AM_Cat, 4 for AF_Cat, 3 for NR_Cat, 3*3*4*3 = 108 patterns in total. Pattern Number=NB_Cat*3*3*4+AM_Cat*3*4+AF_Cat*3+NR_Cat*1+1. Examples data(exampledata) relative.risk(exampledata) risk.summary List the records with relative risks and absolute risks A function to list all the records with relative risks and absolute risks. risk.summary(data, Raw_Ind=1) Arguments data Raw_Ind A data set containing all the required input data needed to perform risk projections, such as initial age, projection age, BrCa relative risk covariates and race. See exampledata for details. The raw file indicator with default value 1. Raw_Ind=1 means RR covariates are in raw/original format. Raw_Ind=0 means RR covariates have already been re-coded to 0, 1, 2 or 3. Value A data frame that includes age, duration of the projection time interval, covariates and the projected risk. A CSV file is created to save the data frame under user s working directory for reading convenience. See Also relative.risk, absolute.risk

risk.summary 15 Examples data(exampledata) risk.summary(exampledata)

Index absolute.risk, 3, 7, 14 BCRA (BCRA-package), 2 BCRA-package, 2 BrCa_1_AR, 4 BrCa_beta, 5 BrCa_lambda1, 5 BrCa_lambda2, 6 check.summary, 7 error.table, 8, 9, 13 error.table.all, 8, 8, 13 exampledata, 9 list.constants, 10 recode.check, 4, 7 9, 11, 14 relative.risk, 4, 7, 13, 14 risk.summary, 14 16